Topics in Cognitive Science最新文献

I present a computational-level model of language production in terms of a combination of information theory and control theory in which words are chosen incrementally in order to maximize communicative value subject to an information-theoretic capacity constraint. The theory generally predicts a tradeoff between ease of production and communicative accuracy. I apply the theory to two cases of apparent availability effects in language production, in which words are selected on the basis of their accessibility to a speaker who has not yet perfectly planned the rest of the utterance. Using corpus data on English relative clause complementizer dropping and experimental data on Mandarin noun classifier choice, I show that the theory reproduces the observed phenomena, providing an alternative account to Uniform Information Density and a promising general model of language production which is tightly linked to emerging theories in computational neuroscience.

Performance on the psychomotor vigilance test (PVT; Dinges & Powell, 1985)-a common index of sustained attention-is affected by the opposing forces of fatigue and sustained effort, where reaction times and error rates typically increase across trials and are sometimes offset by additional efforts deployed toward the end of the task (i.e., an "end-spurt"; cf. Bergum & Klein, 1961). In ACT-R (Adaptive Control of Thought-Rational; Anderson et al., 2004), these influences on task performance have been modeled as latent variables that are inferred from performance (e.g., Jongman, 1998; Veksler & Gunzelmann, 2018) without connections to directly observable variables. We propose the use of frontal gamma (γ) spectral power as a direct measure of vigilant effort and demonstrate its efficacy in modeling performance on the PVT in both the aggregate and in individuals.

Models of the explore-exploit problem have explained how children's decision making is weighed by a bias for information (directed exploration), randomness, and generalization. These behaviors are often tested in domains where a choice to explore (or exploit) is guaranteed to reveal an outcome. An often overlooked but critical component of the assessment of explore-exploit decisions lies in the expected success of taking actions in the first place-and, crucially, how such decisions might be carried out when learning from others. Here, we examine how children consider an informal teacher's beliefs about the child's competence when deciding how difficult a task they want to pursue. We present a simple model of this problem that predicts that while learners should follow the recommendation of an accurate teacher, they should exploit easier games when a teacher overestimates their abilities, and explore harder games when she underestimates them. We tested these predictions in two experiments with adults (Experiment 1) and 6- to 8-year-old children (Experiment 2). In our task, participants' performance on a picture-matching game was either overestimated, underestimated, or accurately represented by a confederate (the "Teacher"), who then presented three new matching games of varying assessed difficulty (too easy, too hard, just right) at varying potential reward (low, medium, high). In line with our model's predictions, we found that both adults and children calibrated their choices to the teacher's representation of their competence. That is, to maximize expected reward, when she underestimated them, participants chose games the teacher evaluated as being too hard for them; when she overestimated them, they chose games she evaluated as being too easy; and when she was accurate, they chose games she assessed as being just right. This work provides insight into the early-emerging ability to calibrate explore-exploit decisions to others' knowledge when learning in informal pedagogical contexts.

Over the past three decades, Van Gelder's dynamical hypothesis has been instrumental in reconceptualizing the ways in which perception-action-cognition unfolds over time and in context. Here, I examine how the dynamical approach has enriched the theoretical understanding of social dynamics within cognitive science, with a particular focus on interpersonal coordination. I frame this review around seven principles in dynamical systems: three that are well-represented in interpersonal coordination research to date (emergent behavior, context-sensitive behavior, and attractors) and four that could be useful opportunities for future growth (hysteresis, sensitivity to initial conditions, equifinality, and reciprocal compensation). In addition to identifying specific promising lines of theoretical inquiry, I focus on the significant potential afforded by computationally intensive science-especially in naturally occurring data or trace data. Building on the foundation laid over the past three decades, I argue that looking to increasingly situated and naturalistic settings (and data) is not only necessary to realize the full commitment to the dynamical hypothesis but is also critical to building parsimonious and principled theories of social phenomena.

Over the past several decades, research in the cognitive sciences has foregrounded the importance of active bodies and their continuous dependence on the changing environment, strengthening the relevance of dynamical models. These models have been steadily developed within the ecological psychology approach to cognition, which arguably contributes to the "ecological turn" we are witnessing today. The embodied and situated nature of cognition, regarded by some as a passing trend, is presently becoming a largely accepted assumption. In this paper, I claim that in light of these developments, ecological psychology, in alliance with related approaches, such as enactivism and interactivism, has the potential to deeply transform our perspectives on cognition and action, restoring their pertinence to humans as persons. However, an important challenge to the realization of this potential has to be noted: neither the mainstream information-processing approach nor the dynamics-oriented perspective on cognition provides an account of how the capacity of humans to use language and think "symbolically" can be derived from the continuous flow of agent-environment interaction. I will attempt to show that posing the "dynamical" and "computational" hypotheses about the nature of cognition as mutually exclusive approaches to cognition results in undesirable reductionism, which makes it difficult to meet this challenge. There are good reasons, advanced over half a century ago by, for example, Michael Polanyi or Howard Pattee, to think that we need complementary descriptions to understand cognizing systems, in order to grasp the fact that they are governed both by physical laws and by emergent historical constraints. Details of such a complementarity-based approach still await elucidation, but some proposed solutions have the potential to ease the tension between the information-processing and dynamical approaches to cognition and to lead to a better understanding of their interrelation.

Artificial intelligence (AI) is often used to predict human behavior, thus potentially posing limitations to individuals' and collectives' freedom to act. AI's most controversial and contested applications range from targeted advertisements to crime prevention, including the suppression of civil disorder. Scholars and civil society watchdogs are discussing the oppressive dangers of AI being used by centralized institutions, like governments or private corporations. Some suggest that AI gives asymmetrical power to governments, compared to their citizens. On the other hand, civil protests often rely on distributed networks of activists without centralized leadership or planning. Civil protests create an adversarial tension between centralized and decentralized intelligence, opening the question of how distributed human networks can collectively adapt and outperform a hostile centralized AI trying to anticipate and control their activities. This paper leverages multi-agent reinforcement learning to simulate dynamics within a human-machine hybrid society. We ask how decentralized intelligent agents can collectively adapt when competing with a centralized predictive algorithm, wherein prediction involves suppressing coordination. In particular, we investigate an adversarial game between a collective of individual learners and a central predictive algorithm, each trained through deep Q-learning. We compare different predictive architectures and showcase conditions in which the adversarial nature of this dynamic pushes each intelligence to increase its behavioral complexity to outperform its counterpart. We further show that a shared predictive algorithm drives decentralized agents to align their behavior. This work sheds light on the totalitarian danger posed by AI and provides evidence that decentrally organized humans can overcome its risks by developing increasingly complex coordination strategies.

The predictive processing framework includes a broad set of ideas, which might be articulated and developed in a variety of ways, concerning how the brain may leverage predictive models when implementing perception, cognition, decision-making, and motor control. This article provides an up-to-date introduction to the two most influential theories within this framework: predictive coding and active inference. The first half of the paper (Sections 2-5) reviews the evolution of predictive coding, from early ideas about efficient coding in the visual system to a more general model encompassing perception, cognition, and motor control. The theory is characterized in terms of the claims it makes at Marr's computational, algorithmic, and implementation levels of description, and the conceptual and mathematical connections between predictive coding, Bayesian inference, and variational free energy (a quantity jointly evaluating model accuracy and complexity) are explored. The second half of the paper (Sections 6-8) turns to recent theories of active inference. Like predictive coding, active inference models assume that perceptual and learning processes minimize variational free energy as a means of approximating Bayesian inference in a biologically plausible manner. However, these models focus primarily on planning and decision-making processes that predictive coding models were not developed to address. Under active inference, an agent evaluates potential plans (action sequences) based on their expected free energy (a quantity that combines anticipated reward and information gain). The agent is assumed to represent the world as a partially observable Markov decision process with discrete time and discrete states. Current research applications of active inference models are described, including a range of simulation work, as well as studies fitting models to empirical data. The paper concludes by considering future research directions that will be important for further development of both models.

An open question regarding how people develop their models of the world is how new candidates are generated for consideration out of infinitely many possibilities. We discuss the role that evolutionary mechanisms play in this process. Specifically, we argue that when it comes to developing a global world model, innovation is necessarily incremental, involving the generation and selection among random local mutations and recombinations of (parts of) one's current model. We argue that, by narrowing and guiding exploration, this feature of cognitive search is what allows human learners to discover better theories, without ever grappling directly with the problem of finding a "global optimum," or best possible world model. We suggest this aspect of cognitive processing works analogously to how blind variation and selection mechanisms drive biological evolution. We propose algorithms developed for program synthesis provide candidate mechanisms for how human minds might achieve this. We discuss objections and implications of this perspective, finally suggesting that a better process-level understanding of how humans incrementally explore compositional theory spaces can shed light on how we think, and provide explanatory traction on fundamental cognitive biases, including anchoring, probability matching, and confirmation bias.